Resilience in Modern Software Systems
summary
Dataset of about 10,000 public incident reports (roughly 75% with a reported duration) shows incident durations are highly skewed so the mean MTTR routinely misleads; Monte Carlo experiments that impose a 10% reduction still produce noisy, often contradictory signals. Resilience is an ongoing capability, not a static ingredient, so instrument socio-technical signals: cost of coordination (people, teams, channels involved), near misses counts, and post-incident participation and linkages to code, docs, and runbooks alongside SLOs/SIs to measure adaptive capacity.
AI-written summary · about
chapters (6)
- 0:00Intro
- 2:31Resilience in 3 acts
- 3:05Act 1: How we talk
- 17:45Act 2: How we measure
- 31:30Act 3: If not MTTR...?
- 37:18Outro
related talks
33mThe Problem with MTTR: Learning from Incident Reports
43mResilience Engineering Mythbusting
29mCulturing Resiliency with Data: A Taxonomy of Outages
44mFrom here to resilience - a travel guide
27mTales from the VOID: The Scary Truth about Incident Metrics
30mResilience Engineering Mythbusting
37mIncident Management Metrics That Matter
05mWhat is resilience engineering